EP1189452A2 - Digital image coding and decoding apparatus and method - Google Patents

Abstract

The discrete cosine transformation of the prediction error is not calculated, but the DCT of the original image to be transmitted and the DCT of the previously reconstructed image are calculated separately from one another. By comparing the amplitudes of the coefficients from the original signal to the difference of the coefficients from the original signal and the reconstructed image, it is decided for each coefficient whether the coefficient of the original signal or the difference of the coefficients is transmitted.

Description

The invention relates to methods and devices for Encoding and decoding a digitized original picture or a digitized coded image.

Such methods and devices are from [1] known.

The well-known, standardized MPEG2 method is used for Compress digital moving picture sequences, that is to say Coding digitized pictures, pixels, which ones usually grouped into picture blocks or picture objects are assigned coding information.

Below is coding information, for example Brightness information (luminance information) and / or Color information (chrominance information) representing the pixels assigned to a digitized image.

The encoding information which is originally present in the local area, according to an MPEG2 iskreten D C osinus ransformation T (DCT) and subjected to the DCT coefficients generated in this way are quantized and entropy coded.

From the quantized DCT coefficients is determined a reconstruction image by means of an inverse quantization and an i nversen D iskreten C osinus T ransformation (IDCT).

In this context, a motion estimation between the original picture and a previous one Reconstruction image carried out.

For coding, i.e. to compress the data to be transmitted Video data is usually between an inter-encoding mode and an intra-coding mode.

As part of the inter-coding mode, the Prediction error that results from the motion estimation as a difference value of the coding information of the coding picture block and the coding information of the image block of the temporally previous picture has been determined, one Subjected to DCT, quantized and entropy encoded fed.

Within the intra-coding mode, without Consideration of coding information temporally previous pictures the coding information of the encoding original image subjected to a DCT and the resulting DCT coefficients are quantized and entropy.

Depending on the available data rate, a rough or a fine quantization of the DCT coefficients chosen, i.e. set.

The coarser the quantization of the DCT coefficients chosen is, the more DCT coefficients become "zero" quantized and the lower is that for the transmission of the encoded original image required data rate.

A is often used to transmit the prediction error lower data rate required than for the transmission of the the intra-coding mode coded DCT coefficients of the coding original image.

However, the motion estimate is inaccurate, which in particular often at the boundaries of moving objects or at block boundaries can be observed between two image blocks, the Data rate for the prediction error can also be higher than the data rate required to encode the original image.

For this reason it is common in known coding methods a switch between the inter-coding mode and the Intra-coding mode provided with which ensures should always be transmitted that signal that overall the lower data rate for its transmission requires.

As a criterion for deciding whether coding according to the inter-coding mode or according to the intra-coding mode is usually used the signal variance. This is directly from the original picture, that is calculated in the local area. Alternatively, it could also by summing the determined DCT coefficients, the means in the frequency domain. The signal with the lower sum of the DCT coefficients has the lower Signal variance.

It is also known that the motion compensating Prediction usually the low-frequency signal components of a video signal is well predicted.

The corresponding DCT coefficients lower frequency Signal components of the respective original image to be encoded therefore have a lower amplitude than the DCT coefficients of the original signal. Owing to Estimation errors in the context of motion estimation are the high-frequency signal components, however, often only with insufficient quality predicted, so the DCT coefficient of the prediction error, which in the following as Prediction coefficients are referred to in this case have a higher amplitude than the DCT coefficients of Original image.

The invention is therefore based on the problem encode or decode digitized image, which a lower one than known coding methods Transmission rate required for its transmission.

The problem is caused by the methods and the devices for coding or decoding a digitized Original picture or a digitized coded picture with the features according to the independent claims solved.

In a method for coding a digitized Original picture with pixels, the coding information is assigned, spectral coefficients of the Coding information of the original image determined. Further become spectral coefficients of a reconstruction image of a previous image determined. For at least part of the determined spectral coefficients are coefficient differences from spectral coefficients of the Original image and the corresponding spectral coefficients of the reconstruction image determined. By forming the respective difference of a spectral coefficient of Original image and the corresponding spectral coefficient of the reconstruction image is a Prediction coefficient formed. The determined Prediction coefficients are matched with Spectral coefficients of the original image compared. The smaller coefficient is at least partially selected and coded.

The invention can clearly be seen in that selectively at the level of abstraction of the individual Spectral coefficients of the original image, which to be encoded is and a reconstructed image of a temporal previous picture, that spectral coefficient is selected and transmitted, its spectral coefficient has a smaller value, that is, a lower value has spectral power density.

It does not become a DCT as in the prior art of the prediction error, but it becomes both a Spectral transformation on the original image to be encoded as also on the reconstructed image of a temporal previous image calculated separately and based on a Comparison of the amplitude of the coefficients of the Original signal to the difference of the coefficients from Original signal and the reconstructed picture can be for everyone Coefficients are decided separately, which in the frame the coding of the picture selected, coded and transmitted becomes. In this case, the recipient will also receive the decision made signals.

In this way, a significant reduction in the required Data rate for coding and transmission of a digitized Image in a moving image sequence.

As coding information, for example Brightness information and / or color information Be assigned to pixels.

According to one embodiment of the invention, it is provided that only the prediction coefficients with the value zero selected and encoded.

A prediction coefficient has a non-zero value according to a further development of the invention appropriate, that is, at the appropriate location within the spectral coefficient of the image Original picture selected and coded.

According to this configuration is a very simple and therefore Coding that can be carried out with very little computing time achieved with improved coding efficiency.

According to a further embodiment of the invention, it is provided for all coefficients of the to be encoded Original image a comparison to the one described above Way to perform and the appropriate coefficient select and encode, using a coefficient suitable is when the selection to a lower amount of data for the Coefficients and the selection information leads.

According to a a further embodiment of the invention, a discrete cosine Transformation can be used, creating DCT coefficients be formed.

To easily decode the encoded picture to enable one for each selection made Selection decision information, for example in the form of a bit, which specifies whether a prediction coefficient or selected a spectral coefficient of the original image and has been encoded, stored and sent to a decoder Transfer decoding.

The selection decision information can, for example, in a selection decision vector of the dimension of transferred coefficients. The Selection decision vector is encoded appropriately and broadcast.

In a method for decoding a digitized encoded image with spectral coefficients, Prediction coefficients and selection decision information, with which is indicated in each case whether the respective coefficient is a Spectral coefficient of the original image or the Prediction error is for a time previous reconstruction image spectral coefficients educated.

The encoded picture is taken into account Spectral coefficients, prediction coefficients and Selection decision information of the coded image and the Spectral coefficients of the previous one Reconstruction image decoded.

Corresponding devices for performing the above described methods each have at least one Processor, which is set up so that the above process steps shown are carried out.

The invention can be done by means of a special electrical circuit, that is, in hardware, as well by means of a computer program, i.e. in software, will be realized.

The invention can be general in any block-based and / or object-based method for coding and decoding a digitized image, especially a sequence of digitized images, i.e. a moving image sequence, be used in which a motion compensating Prediction, especially a motion estimation and a Motion compensation, is used.

An embodiment of the invention is in the figures shown and will be explained in more detail below.

Figur 1

ein Blockdiagramm, in dem eine Vorrichtung zum Codieren einer Bewegtbildsequenz gemäß einem Ausführungsbeispiel der Erfindung dargestellt ist;

Figur 2

ein Blockdiagramm, in dem ein Videokommunikationssystem mit einer Kamera, einem Codierer und einem Decodierer gemäß einem Ausführungsbeispiel der Erfindung dargestellt ist;

Figur 3

ein Blockdiagramm, in dem eine Decodiervorrichtung zum Decodieren eines codierten Videodatenstroms gemäß einem Ausführungsbeispiel der Erfindung dargestellt ist.

Show it

Figure 1

a block diagram showing a device for coding a moving picture sequence according to an embodiment of the invention;

Figure 2

a block diagram showing a video communication system with a camera, an encoder and a decoder according to an embodiment of the invention;

Figure 3

a block diagram illustrating a decoding device for decoding an encoded video data stream according to an embodiment of the invention.

2 shows a video communication system 200 with a camera 201, with which a scene is recorded by means of a sequence 203 of images which follow one another in time.

Scene 202 has any number of different ones Objects on.

The captured sequence 203 of images is taken by the camera 201 via a first communication link 204, for example a cable or a radio link, one Coding device 205 supplied.

Upon receipt of images 203 via an input / output interface 206 by the coding device 205 Images from an analog-to-digital converter 207 of the Coding device 205 digitized and the digitized Images are stored in a memory 208.

The coding device 205 also has a processor 209 on, which is set up so that the further described method steps are feasible.

The logical components that are required in the context of coding or decoding according to the exemplary embodiment are shown in FIGS . 1 and 3 and are explained in detail below.

The functionality of these individual components are according to the embodiment as a program in the memory 208 of FIG Coding device or in a memory in the further described decoding device and are stored by the processor 209 of the coding device 205 or the Decoding device, as explained in more detail below, executed.

The individual components of the coding device 205 are over a computer bus 210 coupled together.

The coding device 205 is via the input / output interface 206 and over a second Communication link 211, for example a Radio connection or a landline, that is Cable coupled to a decoder 212.

Via the second communication link 211 digitized, encoded video images 213 from the Coding device to the decoding device 212 transfer.

The coded pictures 213, 214 are of the Decoder 212 via an input / output interface 215 received and in a memory 216 saved.

Furthermore, the decoding device 212 has a processor 217, which is set up in such a way that the corresponding Method steps for decoding the coded pictures and thus performed to reconstruct digitized images can be.

The components of decoder 212 are one Computer bus 218 coupled together.

1 shows the logical structure of the coding device 100, which according to this exemplary embodiment is implemented in the form of a computer program which is carried out by the processor 207.

The coding device 100 can, however, in its components also each as a special electronic circuit be implemented on a separate computer card.

The pixels grouped into image blocks are called digitized image 101 supplied to the coding device 100.

In a transformation unit 102, a D is the image blocks iskrete C osinus ransformation T (DCT) is performed.

Furthermore, the pixels of the to be encoded Original image 101 of a unit for motion estimation and Motion compensation 110 supplied.

The DCT coefficients formed by the transformation unit 102 106 are stored in a memory 103.

Furthermore, the DCT coefficients of the original image 101 read out from the memory 103 and as read DCT coefficients 107 to a subtraction unit 104.

In the subtraction unit 104 is read from a Spectral coefficients 107 of the original image to be encoded 101 a corresponding one from the unit to Motion estimation and motion compensation 110 educated Spectral coefficient 109 of the temporally preceding image, which corresponds to the corresponding image position.

The corresponding one by forming the difference Spectral coefficients of the original image 101 to be encoded and the difference determined in the previous image, hereinafter referred to as difference coefficient 108 fed to a quantization unit 113 and there quantized.

Furthermore, the spectral coefficients 107 of the coding original image 101 also the Quantization unit 113 supplied and quantized there.

The quantized coefficients 107, 108 become one Decision unit 105 fed, for example one Unit for forming the maximum 105 that of the unit 105 for Formation of the maximum input variables.

The respectively larger value from the spectral coefficient 107 and the coefficient difference 108 is selected Coefficient 111 selected and encoded.

Alternatively or additionally, the selection can be made in the Decision unit 105 the data rate for the signaling are taken into account.

Furthermore, the respective selected quantized Coefficient 115 and an indication 112 as to whether it is the quantized coefficient by one Spectral coefficients 107 of the original image or around Difference coefficient 108 (hereinafter referred to as coefficient 112), a unit for Entropy coding 116 supplied, thereby encoding Video data stream 121 is formed, which to the Decoding device 300, which will be described below, is transmitted.

In a feedback loop 114, the selected and quantized coefficients 115 and the coefficient specification 112 to an inverse quantization unit 117 and converted there to inversely quantized coefficients 118.

The inverse-quantized coefficients 118 are transformed in an inverse transformation unit 119 by an inverse d iskreten C osinus T ransformation (IDCT) for inverse-coded coefficients 120th

The inverse coded coefficients 120 become the unit Motion estimation and motion compensation 110 supplied, in which a usual motion estimation or Motion compensation takes place.

According to this procedure, everyone gets to transmitting DCT coefficient checks whether the DCT coefficient of the original image 101 or the reconstructed one Image, i.e. the prediction error image depending on which coefficient is smaller, i.e. has a lower value.

3 shows the decoding device 300 to which the encoded video data stream 121 is supplied.

In an inverse quantization unit 301, inverse quantized spectral coefficients 302 formed, one inverse transformation unit 303 is supplied in the an inverse discrete cosine transformation to the inverse quantized coefficient 302 is performed for formation inversely transformed coefficient 305.

The inversely transformed coefficients 305 and the specification about whether the coefficients are Difference coefficients or spectral coefficients of the An original image, an adder 304 fed.

Accordingly, either one described below reconstructed DCT coefficient 311 of the temporal added to the previous image or the reconstructed one Coefficient 305 is simply referred to as spectral coefficient 306 in stored in a memory 307.

In the motion compensation unit 308 for the In the event that the coefficient 305 is one Difference coefficient 108 is one Motion compensation performed and that motion compensated signal 309 becomes one Transformation unit 310 for performing a DCT transformation fed.

The DCT coefficients formed are generated by the transformation unit 310 311 of the reconstructed image of Addition unit 304 supplied and in the event that it is is a difference coefficient 108, the respective DCT coefficient formed by the transformation unit 310 311 to the inverse quantized coefficient 302 added.

The memory 307 is used to store what is stored there reconstructed image as reconstructed video signal 312 read out and the user of the decoding device 300 for example displayed on a screen.

The following publication is cited in this document:

[1] J. De Lameillieure, R. Schäfer, MPEG-2 image coding for digital television, television and cinema technology, 48th year, No. 3/1994, pp. 99 - 107, 1994

Claims (10)

Method for coding a digitized original image with pixels to which coding information is assigned where the spectral coefficients of the original image are determined, in which spectral coefficients of a reconstruction image of a previous image are determined, in which coefficient differences are determined for at least some of the spectral coefficients determined from spectral coefficients of the original image and the corresponding spectral coefficients of the reconstruction image, thereby forming prediction coefficients, in which the determined prediction coefficients are compared with corresponding spectral coefficients of the original image, and in which the smaller coefficient is at least partially selected and encoded. Method according to claim 1, in which only the quantized prediction coefficients with the value zero are selected and coded, and in which the spectral coefficients of the original image are otherwise selected and encoded, or vice versa. The method of claim 1 or 2,
in which brightness information and / or color information is assigned to the pixels as coding information. Method according to one of claims 1 to 3,
in which the smaller coefficient is selected and encoded for all coefficients of the original image to be encoded. Method according to one of claims 1 to 4,
in which DCT coefficients or integer transformation coefficients are formed as spectral coefficients. Method according to one of claims 1 to 5,
in which a selection decision information is stored for each selection of a coefficient and transmitted to a decoder. Method for decoding a digitized coded image with spectral coefficients, prediction coefficients and selection decision information, each of which specifies whether the respective coefficient is a spectral coefficient of the original image or the prediction error image, in which spectral coefficients are formed for a temporally preceding reconstruction image, and in which the encoded image is decoded taking into account the spectral coefficients, prediction coefficients and selection decision information of the encoded image and the spectral coefficients of the previous reconstruction image. Device for coding a digitized original image with pixels to which coding information is assigned, with a transformation unit, for determining spectral coefficients of the original image or a reconstruction image of a previous image, with a subtraction unit for forming coefficient differences from spectral coefficients of the original image and the corresponding spectral coefficients of the reconstruction image, as a result of which prediction coefficients are formed, with a comparison unit for forming a comparison value between the prediction coefficients and corresponding spectral coefficients of the original image, and a coefficient selection unit for selecting the appropriate coefficient, and an encoding unit for encoding the selected coefficients. Device for coding a digitized original image with pixels to which coding information is assigned,
with a processor that is set up in such a way that the following steps can be carried out: Determining spectral coefficients of the original image, Determining spectral coefficients of a reconstruction image of a previous image, Determining coefficient differences for at least some of the determined spectral coefficients from spectral coefficients of the original image and the corresponding spectral coefficients of the reconstruction image, thereby forming prediction coefficients, Comparing the determined prediction coefficients with corresponding spectral coefficients of the original image, and at least partially selecting and coding the respective smaller coefficient. Device for decoding a digitized coded image with spectral coefficients, prediction coefficients and selection decision information, each of which is used to indicate whether the respective coefficient is a spectral coefficient of the original image or of the prediction error image,
with a processor that is set up in such a way that the following steps can be carried out: Spectral coefficients are formed for a previous reconstruction image, the encoded image is decoded taking into account the spectral coefficients, prediction coefficients and selection decision information of the encoded image and the spectral coefficients of the reconstruction image preceding in time.

Images